Abhishek Sanjay Sonawane

Abhishek Sanjay Sonawane

Computer Science Graduate Student at Arizona State University

Academic Projects

Disaster response portal

Javascript, React, D3.js, HTML, CSS, Bootstrap

Dr. Chris Bryan, Arizona State University

  • Developed a portal to help emergency responders in planning rescue operations during the time of crisis.
  • Crowd-sourced data collected from a disaster reporting mobile application was visualized in the form of line chart, choropleth map, dot plot, scatter plot, parallel co-ordinate plot, box plot and heatmap.
  • The plots were interlinked to provide the users a cohesive understanding of the situation on ground, changes in the conditions over time and assess the reliability of the data reported.
  • The failure of a parallel co-ordinate plot in visualizing large number of records was overcome by innovating a visualization technique which blended box-plot along with a parallel co-ordinate chart. This was done by grouping the records based on a meaningful criteria, in this case region where reports were made, and using a box-plot for each axis on parallel co-ordinate plot to represent the distribution of values within each group.

Galaxy Classification using Transfer Learning

Python, Keras, TensorFlow, pandas, NumPy, Matplotlib, scikit-learn

Dr. Hannah Kerner, Arizona State University

  • The shape of a galaxy provides insights about its evolution and is an important aspect in astronomical research.
  • Novel deep learning classifiers were proposed and trained to categorize galaxies into 10 classes using the galaxy images captured by space exploration missions.
  • Transfer learning was used as opposed to the earlier works that trained networks from scratch.
  • VGG-16, ResNet-50, XceptionNet, InceptionNet, and DenseNet-121 architectures were used to create 5 separate feature extraction models.
  • The VGG-16 based model turned out to be the best one outperforming a state of the art model in 8 of the 10 classes giving only 2% lower overall accuracy.
  • The resource consumption in training the model was brought down by 60% and the number of models trained before finding the one that gives the best performance were reduced from several 100s to 10.

Autonomous Warehouse

Clingo, Python

Prof. Ali Altunkaya, Arizona State University

  • Developed a program that prepares a schedule of operations to be performed by robots in a warehouse to fulfill all customer orders in minimum possible time.
  • Declarative programming paradigm was used since this was a search space optimization problem.

Processing loan applications using Machine Learning

Python, NumPy, Pandas, Matplotlib, scikit-learn

Dr. Yoo-Jung Choi, Arizona State University

  • Compared the performance of 5 different ML classifiers : logistic regression, decision tree, support vector machine, k-Nearest Neighbors (kNN) and gaussian naiver bayes, for predicting whether a loan application must be approved or not.
  • The use of effective data-preprocessing techniques for dealing with missing values and removing non-essential features ensured that these models outperformed prior work done in this field.
  • Decision tree model gave perfect accuracy of 100% and was suggested to be the best classifier for this application.

Optimizing the running time of KNN classifier using parallelism

C, C++, OpenMP

Prof. R. A. Kulkarni, Pune Institute of Computer Technology

  • Implemented KNN algorithm for performing skin segmentation on images.
  • The time required for classification was reduced up to 66% by using parallel processing without affecting the classification performance.

Accessing cloud storage using POSIX file access semantics

Golang, C, C++, Amazon Web Services, Google Cloud Platform, MinIO

Dr. A. R. Buchade, Pune Institute of Computer Technology

  • Cloud platforms offer several advantages over on-premise data storage. However, many people are still unfamiliar with the semantics of accessing cloud storage and several softwares rely on the old POSIX semantics for data access. This project provided access to cloud storage using POSIX semantics.
  • Plugins were developed to facilitate data transfer between a Server Message Block (SMB) protocol based file server and cloud storage platforms like AWS and GCP.
  • Although the data saved on file server was stored on cloud, the users were able to access it like any other file or folder stored on their device.

Mitigation of cyber-attacks on web applications

Java, JSP, Servlet, MySQL, HTML, CSS

Prof. Preeti Jain, Pune Institute of Computer Technology

  • Proposed and implemented the use of a policy-based proxy server to tackle SQL injection (SQLi) and Cross Site Scripting (XSS) attacks on web applications.
  • Developed two versions of a web application for banking purposes. One version made use of a proxy server while the other version didn't.
  • The proxy server receives and analyzes each request received from the client. It consists of a set of policies that determine whether a request is likely to cause an SQLi or XSS attack.
  • If a request is likely to cause an attack, it is either sanitized or terminated before sending it to the main server.
  • In the absence of a proxy server, the requests are directly sent to the actual server.
  • The two versions of the application were used to demonstrate teh effectiveness of the proxy server in tackling different types of attacks.

Object detection for the vision impaired

Python, Raspberry Pi, TensorFlow, OpenCV, HTML, CSS, PHP

Dr. A. R. Buchade, Pune Institute of Computer Technology

  • An autonomous system to help vision impaired people get an understanding of the objects in their environment.
  • Single Shot MultiBox Detector was used to find the names of objects present in the live images captured by Raspberry Pi Camera module.
  • The list of these names was made available on a website which had text-to-speech facility for accessibility purposes.

Grocery Store Management System

Java, Swing, MySQL, Sockets, Multi-threading

Prof. Pujashree Vidap, Pune Institute of Computer Technology

  • Desktop applications for the daily activities at a grocery store.
  • Developed two separate applications one for the clients who shop at the grocery store and the other for vendor.
  • The vendor's application acts as a server and multiple clients can be logged in through different computers at the same time.
  • Multi-threaded server was developed to handle requests from multiple clients at the same time.
  • Username and password were used for authenticating customers and the vendor. Different applications would be opened depending on whether a client or a vendor has logged in.
  • Sign-up page was implemented for creating the profile of new users.
  • The profile of a user consists of identification information like name and address, a profile picture, and the signature which could be drawn by dragging the cursor in a rectangular region on sign-up page.
  • Profile information including the photo and signature of each user gets saved in a MySQL database running on server.
  • All items available for the customers were distributed into different categories like vegetables, fruits, cosmetics, etc. and a separate page was displayed whenever the user clicked on a category.
  • A cart was provided to keep track of the products selected by the customer.
  • When users check out, they get charged for items in their cart and an invoice is generated.
  • The vendor's application keeps track of the number of items available in the inventory. The quantities of each item are updated as and when they are purchased by customers.
  • New items can be ordered by the vendor through the application and inventory information will be updated accordingly.
  • History of all transactions is saved on the server side.